Advancing Personal Reading Stats — Startup Week 22
Kaffae has started with a simple analytics dashboard to understand your behavior. It showed timeline graphs, piecharts, and wordcloud all based on your reading. It had a great reception at the time.
But that was a year ago already. My long gone 2019 commit shows the story sadly. Since then I’ve been adding different features such as suggestions, social feed, and reflection report. It’s not that I didn’t care about the dashboard page. It’s just that I was busy expanding the functionality of this app.
This week that’s changed. I finally made a major update on the analytics dashboard, now being called as advanced statistics.
What Does It Mean For Users?
Users get much more value out of the reading graphics now.
Focus Reading
I added additional metrics in the read count timeline. Many people skim articles. The ratio is about 8:2. And I’ve been getting feedback that people wanted to separate the two more vividly.
With this addition, it’s clear the proportion of your skimmed articles vs long read articles now. You would want to increase the ratio of long reading.
Lengthy Count
Unlike books, there is no consistency or an accepted standard in how long the articles should be. But most people would agree, very short articles as clickbaity and unworthy.
I added an additional graph that divides your articles in how long. If you always read short articles, there’s a high chance you are suffering from clickbait reading. Or you might decide you want to challenge more challenging materials.
Word Complexity
How difficult is the type of articles you are reading? This chart is best accompanied by the previous lengthy graph. It will show the type of writings used in the articles.
Before the graph was based on word difficulty picked based on the metrics used to categorize books into different suitable grades. But that had limited functionality. It didn’t capture the full complexity of ideas expressed in writings. What people wanted to know is how in-depth your knowledge is compared to other people. That’s why I made a measurement based on word frequency from the entire database. Common words such as dogs or cats appear most often. An example of rare words is adversarial network (one of the deep learning techniques). It checks the proportion of rare and common words in your articles and shows how niche the topic of your reading is.
Better Coverage in Political Lean
If you are a news guy, you’d like to check what type of bias happens. The concept dates back from the previous presidential campaign about the mention of political bubble and polarization. As I’ve tested this feature myself, we’d be much better off reading left and right media. The problem is we don’t know if our favorite media falls onto which political flavor.
The selection is based on Media Bias Check website. Before it only covered 40 domains. And now, that has been extended to all that’s been listed on the site. That amounts to more than 400. It is the coverage increase of 10 times.
Helpful Tips
Just by having graphs, users were left confused about the meaning of each graph. Now, the app has prompt titles with your perspective: for example “what is your favorite length of reading?”
Also, I added a tooltip hover popup that explains the calculation mechanism for each graph. For example, I’ve had questions around sentiment breakdown, especially from those who were concerned that negative sentiments equate bad content. I added a statement explaining what types of words that are used in calculating the sentiment; words such as success, helpful, and innovation are considered positive sentiment.
Refined Breakdown
Refined breakdown helps you narrow down the reflection of your own reading.
There were various piecharts from political lean to word complexity. But they had only 3 breakdowns. I turned them all to a more refined 5. Five was a good balance between oversimplicity and messiness. It heightens the top category well, which only captures the selected few now.
Aesthetic Improvement
The first iteration used the default color schema from Bootstrap.
Now, it tailored to suit the branding of the app. It has a visually consistent feeling.
Advanced statistics will provide the value very upfront. This information is only unique in Kaffae, and upon feedback, it will continue to be improved.
